Senior AI engineering · Fixed-price · Overlaps US Eastern & Central

Turn your AI demo into a feature customers trust.

We build AI assistants over your docs, agents that take actions, and AI features inside your SaaS — tested for quality, tracked for cost, and delivered at a fixed price. You work directly with the senior engineer who builds it.

Book a free 30-min call → See the Orbiill case study

Julian Lopez · Founder & lead AI engineer

Orbiill · Billing AI

ORBIILL

Billing AI

Dashboard

Pricing

Subscriptions

Settings

Account

Stripe Connection

Workspace Production

✦ 5/5 Jackson Hill

← Pricing

Prompt 1 Draft

v28 · Created Jun 11, 2026

Plans 1

Most AI features die between the demo and production.

It wowed the demo — then hallucinated in front of a customer.

A prototype that works once isn't a system you can put in front of real users.

Token costs looked fine in testing, then exploded at real volume.

Without cost tracking and caching, inference spend grows faster than your usage.

An agency shipped a prototype and left — no evals, no docs.

No way to measure quality, no way to improve it safely, no one who understands the code.

We build the version that actually ships — tested for accuracy, cost-tracked, and stable under real traffic.

// our work

What we've shipped.

A client case in production, Julian's experience before Drembyte, and an open build with public code.

Orbiill Assistant Live

How many configs does Free include?

SOURCES

Free, Pro & quotas FAQ

The Free plan includes one active pricing config and 5 AI generations / month.

Client work · illustrative recreation, no client data

Orbiill: a cited AI assistant inside a SaaS billing product

Orbiill's users needed fast, correct answers about pricing configs, plans and Stripe setup — without digging through docs or waiting on support.

What we builtAn assistant inside the product that answers from Orbiill's own docs, shows the source of every answer, and knows each customer's setup.
How we checked qualityA list of real customer questions, re-tested before every change, so every change is measured before it ships.
What shipped and what the client saidTaken from proof of concept to production and integrated into the product. "Significantly leveled up our user experience." — William Johnson, Founder

Read the case study →

Before Drembyte

Julian's work as an employee — shown as experience, not as studio clients.

Founder experience · employer under NDA

AI engineering at a US EdTech platform

Three years helping build AI features for a platform with 1M+ learners: AI search over learning content, AI agents that work with the platform's data, and fast, streaming AI answers with costs tracked per request.

1M+learners reached

3 yrsbuilding AI there

Founder experience · contractor

GoDaddy, via Altimetrik

As an engineer at Altimetrik, Julian helped build GoDaddy's shared frontend platform — the habits that carry into every AI project: safe rollouts and code other teams can maintain.

Open build

A reference project with public code, built to test the patterns we use in client work.

Open build · source on GitHub

AI support agent

A support assistant that searches a help center, reads and opens tickets, understands screenshots and PDFs, and tracks what every answer costs — while the reply streams in.

Read the code →

// services

What we build — priced up front, scoped in writing.

Every engagement starts with a fixed scope and price, agreed in writing. Prices below are introductory founding-client rates.

Start here

AI Production Audit

Your AI feature gives wrong answers, costs too much, or nobody can tell whether it's getting better. In five business days, you receive a clear diagnosis and a fixed-price plan to fix it.

Book a call about an audit →

A quality scorecard on your real user questions
Cost and speed findings, per request
The highest-impact fixes + a fixed-price build plan

From $1,5005 business daysFee credited if you start the build within 60 daysSee what's in the report →

RAG development & AI search

AI search & assistants over your docs

Your customers ask, your docs answer — correctly, with sources.

Support assistant over your help center
Internal assistant over policies, contracts and wikis
Every answer cites its source, tested before launch

From $3,000Typically 3–5 weeks

How we build RAG →

AI agent development

AI agents that do real work

Agents that take actions in your tools — with guardrails and a human in the loop.

Support agents that resolve tickets, not just reply
Lead qualification and routing
Data extraction from PDFs, invoices and forms

From $5,000Typically 4–6 weeks

How we build agents →

AI feature integration for SaaS

Add AI features to your SaaS

Summaries, generation and copilots inside your product — streamed, cost-tracked, rolled out safely.

In-app copilots and smart actions
Cost tracked per request, kept low with caching
Gradual rollout you can switch off instantly

From $4,000Typically 3–5 weeks

How we add AI to SaaS →

MCP server development

Make your product agent-ready

Let your customers use your product from Claude, ChatGPT and other AI assistants — safely, through an MCP server (the standard way AI assistants connect to apps).

Your product's data and actions, usable by AI assistants
Access limited to exactly what you allow
Docs your customers can plug in on day one

From $2,000Typically 1–2 weeks

How we build MCP servers →

Founding-client pricing · up to three portfolio partnerships

Introductory rates for our first clients — in exchange for a case study.

Drembyte is a young studio, and we're building our portfolio deliberately. Our first clients get the introductory rates above. In return, if the engagement produces useful results, we ask for permission to publish a short case study and a testimonial. Ask whether a partnership is still available.

Ask about founding-client rates → No commitment — the first call is free.

// what you get

What lands in your repo, every engagement.

Not a demo and a goodbye. These are the artifacts your team keeps.

Quality report

Accuracy tested on your real questions, with scores before and after every change.

Cost dashboard

Dollars per request and per feature, so spend never surprises you at scale.

Code, docs & handoff

Reviewed, tested code in your repo, architecture docs, and a walkthrough so your team owns it.

// how it works

From call to shipped. Evals, not vibes.

We start small, prove it works, and only scale once you've seen results.

Discovery call

30 free minutes to understand your problem, your stack and what success looks like.

Scoped proposal

Fixed scope, timeline and price in writing — whether we move forward or not.

Build & ship

Weekly demos, small safe releases, and tests that measure quality.

Handoff or retain

Full code and docs handoff, or a monthly retainer for optimization and new features.

A written plan before you commit

If it's a fit, after the free call you get a concrete build plan with scope, timeline and a fixed price — yours to keep whether we move forward or not. If the problem needs deeper digging first, we start with an audit.

Milestones with written acceptance criteria

Build milestones have acceptance criteria agreed in writing — for assistants and agents, a quality target on real questions. If a milestone misses the agreed bar, payment is deferred until those criteria are met.

// who you'll work with

Meet the engineer who builds it.

I'm Julian Lopez, a senior AI engineer and the founder of Drembyte. For the last three years I've helped build AI features for a US EdTech platform serving over a million learners — RAG pipelines on pgvector, multi-agent systems across Claude, OpenAI and Gemini, custom MCP servers, and streaming UIs with per-request cost tracking. Before that, as an engineer at Altimetrik, I helped build GoDaddy's shared frontend platform.

Drembyte is small on purpose. I lead every engagement and write the critical code myself, with support from a small team when a project needs it. No account managers, no junior hand-offs — you talk to the person building your system.

— Julian

TypeScriptNext.jsNestJSSupabase + pgvectorClaude APIOpenAILangGraphMCPAWS

// working together

Built for US teams. Easy to work with.

Based in Medellín, Colombia (UTC-5). We work in your repository, overlap with US Eastern and Central hours, invoice in USD, and hand off code and documentation under a written agreement.

You own the code

Ownership of all code and IP transfers to you under our written agreement. Work lives in your repo from day one.

NDA first

We sign your NDA before you share anything sensitive. Clear written scope for every project.

Invoiced in USD

Paid per milestone, never one big upfront check.

Your business hours

A full working day of overlap with US Eastern and Central; async-friendly for the West Coast.

Fast replies

Replies within one business day, by email, WhatsApp or your team chat.

Your data stays yours

Least-privilege access, secrets out of code, model-provider data settings configured and documented, and compliance needs scoped up front.

// what clients say

“Working with Julian Lopez to develop the AI Chat features for orbiill.com was a seamless experience. He successfully implemented a powerful RAG system that has significantly leveled up our user experience. Professional, highly skilled, and efficient — highly recommended!”

William JohnsonFounder, Orbiill · on the AI assistant in the case study

// is this a fit?

Let's be honest about fit.

A great fit if —

You have a real product with real users.
You need an AI feature that works, not a science experiment.
You want to talk directly to the engineer who builds it.

Probably not if —

You need a 10-person team or staff augmentation.
You're shopping for the cheapest possible quote.
You need 24/7 enterprise SLAs today.

Questions, answered.

How long does a typical project take?

Most builds take 3–6 weeks: an MCP server is often 1–2 weeks, a production RAG assistant 3–5, an agent with several tools 4–6. You get the exact timeline in the scoped proposal before you commit.

How do contracts and payments work for US companies?

A written proposal with fixed scope and price, your NDA if you have one, and invoices in USD paid per milestone. No large upfront payment and no open-ended retainers without agreement.

Who owns the code?

You do. Ownership of all code and IP transfers to you under our written agreement, and the work lives in your repository from day one. You also keep the quality tests, docs and runbooks.

Why a small studio instead of an agency?

You talk to the person who writes the code — no account managers, no junior hand-offs. We work in your repos and tools, join your code reviews, and leave your engineers able to own the system.

I only have an idea — is that enough to start?

Yes. Many projects start as "we think AI could help here." The discovery call — and if needed, an audit — turns that into a concrete, scoped plan before you commit to a build.

How do you handle our data and security?

Least-privilege access, secrets out of code, and an NDA before anything sensitive. For regulated data we scope compliance requirements up front.

Ready to ship AI that actually works?

Book a free 30-minute call. You'll get a straight answer — and, if it's a fit, a concrete fixed-price plan. Sometimes the answer is "you don't need us."

Book a free 30-min call → Send a project brief instead

Pick a time

30 minutes, free, no prep needed. Times are shown in your time zone.

Open the calendar →

Prefer to write?

Email or WhatsApp us directly — or use the form.

emailjulian.lopez@drembyte.com
whatsapp (US number)+1 (310) 690-3744
hoursMon–Fri, US Central / Eastern hours

Send a project brief

A few lines is plenty — we reply within one business day.